The use of artificial neural networks in discriminating partial discharge patterns

A novel alternative to statistical analysis is the use of artificial neural networks (ANNs). This paper investigates their use in recognising the PD patterns of solid-insulation test samples which contain a different number of cylindrical artificial voids. One of the aims is to determine whether such a technique is sensitive enough to detect the slight difference between these patterns. A typical three-layer network structure with feedforward connections is chosen together with the back-propagation learning method. The network used is also more complex with four output neurodes. Techniques to accelerate the training process of the neural network are also discussed